Hi,
I also have similar observations. I measured several printed antennas in an anechoic chamber. In some cases, the results were close but in other cases, there was a huge difference. HFSS gave a 94% efficiency of a printed monopole designed on FR4 substrate. However, when measured in Satimo anechoic chamber, the actual efficiency was only 56%. For the patch antenna, the measured readings were mostly the same.
If you get to know how to get accurate efficiency with HFSS, please share.
Also there is a Wheeler's cap method to test the antenna efficiency using network analyzer. I never get the correct reading using that method. If anyone knows about this method, please share.